Eric Schmidt on AI, the Battle with China, and the Future of America

The Quiet Surge: How One Technologist Sees a New Intelligence Shaping Power

Eric Schmidt’s conversation moves like a tour through a world reconfigured by machines, rockets, capital flows, and geopolitics. The through-line is not a single breakthrough but a set of converging shifts: more capable software agents, divergent development models between countries, a reimagined battlefield, and an industrial appetite for difficult hardware projects like rockets. The thread that ties them together is a conviction that intelligence—nonhuman, networked, and increasingly agentic—will alter how nations compete, companies innovate, and societies allocate risk and reward.

Agents, Volition, and the Middle-to-Middle Work of Machines

Schmidt frames the present as an agentic revolution: collections of software modules that can act, plan, and collaborate autonomously. These agents are not metaphors but tactical tools that will reshape everyday tasks and business processes. He underscores an important distinction: current systems are not end-to-end replacements for humans. Instead, they operate in the middle—generating drafts, running simulations, and iterating on problems while humans supply goals, context, and moral judgment.

Why that matters

That gap between middle and end is more than technical pedantry; it defines the next decade of labor, product design, and governance. When machines do the heavy lifting of reasoning but still require human objective-setting, firms that design robust human-in-the-loop workflows will extract disproportionate value. The more these agents can autonomously coordinate—forming ad hoc coalitions to solve problems—the less predictable the strategic landscape becomes.

A Divided Architecture: Open Weights, Closed Data, and the Global Split

One of the most consequential distinctions Schmidt raises is structural: a divergence in how models are shared. Western firms tend toward closed weights and proprietary data; some Chinese actors push open weights and openly shared training sets. The practical effect is geopolitical: nations and companies will choose model ecosystems that align with their infrastructure, regulatory preferences, and commercial interests. The result could be an emergent technical bifurcation—different language models, different defaults, and different norms around intent and safety.

• Open models spread quickly across markets with looser capital constraints and aggressive deployment priorities.
• Closed models concentrate advanced research in deep-pocketed labs that can underwrite expensive compute and controlled releases.

Hardware, Capital, and the Geography of Capability

Schmidt rejects a simple race narrative in favor of a layered account. China’s strategy, he says, emphasizes relentless application—embedding AI into consumer apps, robots, and operational systems—while U.S. efforts tilt toward the abstraction of general intelligence research. That orientation is shaped by capital markets and compute access: large-scale AGI work depends on enormous datacenter investment, while applied systems can proliferate on more modest budgets.

The implication is practical: dominance in AI will not be determined by a single metric but by a portfolio of strengths—compute capacity, venture capital depth, industrial know-how, and the ability to ship reliable products at scale.

Rockets, Startups, and the Persistence of Hard Things

Relativity Space, the company Schmidt recently joined as an investor and chair, serves as a necessary reminder that the future is not purely virtual. Rockets are stubbornly physical: most of a launch vehicle is propellant, and marginal gains are measured in engineering creativity as much as in beautiful code. The space economy remains a domain where hardware complexity, manufacturing endurance, and patient capital intersect. Betting on difficult, hardware‑intensive industries is a bet on durable competitive advantages and on belief in the industrial capacity to iterate and scale.

How War Changes When Software Leads

Perhaps the most unnerving section of the discussion concerns how automation remakes conflict. Schmidt describes drone swarms, low-cost expendable systems, and reinforcement-learning planning as elements that change both the tactics and the psychology of war. Cheap drones alter economics; sophisticated planners make actions less predictable. His core argument is paradoxical but cleareyed: as autonomous systems proliferate, deterrence dynamics shift because neither side can reliably infer the other’s plan. That opacity could raise the bar against kinetic escalation, or it could produce catastrophic miscalculation.

A new doctrine emerges

In this future, the forward line is software and sensors; humans are positioned behind algorithms that command mobility and targeting. The result is an arms ecosystem where synthetic training data, iterative learning, and rapid prototyping matter more than traditional inventories of tanks and ships.

Societal Strain: Work Habits, Populations, and the Fragile West

Schmidt also circled back to softer but no less urgent risks: demographics and civic vitality. Declining birth rates, shifting work-life norms, and internal social debates all shape a country’s capacity to sustain growth. He links these trends to national competitiveness—fewer people, older populations, and shrinking domestic markets create an economic drag that is hard to offset even with innovation.

Immigration, he argues, is a blunt but practical response; investing in next-generation education and infrastructure is another. The underlying point is structural: technology matters, but so do the underlying social systems that supply talent, capital, and collective will.

When Does Machine Problem-Solving Become True Generality?

Finally, Schmidt tackles the threshold question of general intelligence. He acknowledges rapid progress—agents, better reasoning on narrow tasks, and enormous gains in specialized domains—but draws a firm line at volition. Current models do not set their own objective functions; they do not exhibit the non-stationary, cross-domain creativity that characterizes human genius. That barrier, he suggests, is the central technical and ethical hinge: if solved, it would transform everything; until then, the pragmatic choreography of humans and tools will determine outcomes.

In the end, the future Schmidt sketches is neither utopia nor apocalypse but a contested landscape: new capabilities that magnify power, markets that redistribute who can build and deploy them, and institutions that must learn fast enough to keep up. The question left behind is not whether machines will get smarter, but whether human systems—political, financial, and moral—can be equally adaptive to steward that intelligence into enduring advantage rather than brittle risk.